Update train.py
Browse files
train.py
CHANGED
@@ -1,26 +1,19 @@
|
|
|
|
|
|
1 |
from datasets import load_dataset
|
2 |
-
from transformers import FastSpeechForConditionalGeneration, Trainer, TrainingArguments
|
3 |
|
4 |
# تحميل البيانات للهجة النجدية
|
5 |
dataset = load_dataset("m6011/sada2022")
|
6 |
najdi_data = dataset.filter(lambda example: example['SpeakerDialect'] == 'Najdi')
|
7 |
|
8 |
-
# إعداد
|
9 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
|
11 |
-
#
|
12 |
-
|
13 |
-
output_dir="./results",
|
14 |
-
per_device_train_batch_size=4,
|
15 |
-
num_train_epochs=5,
|
16 |
-
)
|
17 |
-
|
18 |
-
trainer = Trainer(
|
19 |
-
model=model,
|
20 |
-
args=training_args,
|
21 |
-
train_dataset=najdi_data['train'],
|
22 |
-
eval_dataset=najdi_data['test']
|
23 |
-
)
|
24 |
-
|
25 |
-
# بدء التدريب
|
26 |
-
trainer.train()
|
|
|
1 |
+
from espnet2.bin.tts_train import train
|
2 |
+
from espnet2.tasks.tts import TTSTask
|
3 |
from datasets import load_dataset
|
|
|
4 |
|
5 |
# تحميل البيانات للهجة النجدية
|
6 |
dataset = load_dataset("m6011/sada2022")
|
7 |
najdi_data = dataset.filter(lambda example: example['SpeakerDialect'] == 'Najdi')
|
8 |
|
9 |
+
# إعداد التدريب
|
10 |
+
train_config = {
|
11 |
+
'output_dir': './results',
|
12 |
+
'train_data_path_and_name_and_type': najdi_data['train'],
|
13 |
+
'valid_data_path_and_name_and_type': najdi_data['test'],
|
14 |
+
'train_batch_size': 8,
|
15 |
+
'epochs': 10,
|
16 |
+
}
|
17 |
|
18 |
+
# بدء عملية التدريب باستخدام ESPnet
|
19 |
+
TTSTask.main(**train_config)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|